Machine Learning Driven Energy Optimization in Smart Buildings with Integrated Renewable Energy Sources

Building energy Applications of energy management Building energy use is a large part of the worldwide electricity use and requires smart and effective energy use plans. With the growing accessibility to the high-resolution consumption information provided by buildings with smart technology, it has become possible to implement machine learning (ML) to optimize energy use based on data. The paper shows the design of an hourly energy consumption forecasting and optimization framework of smart buildings using a realworld building power consumption dataset and machine learning. The following methodology implies extensive data preprocessing, temporal Sandboxing of features, and creation of various models of ML, including linear regression, tree-based ensemble procedures, and long short-term memory (LSTM) networks. Even the standard regression measures of mean absolute error, root mean square error, and coefficient of determination are used as measures of model performance. Experimental findings prove that the dynamics of the LSTM model are significantly superior to the conventional methods and the ensemble methods; it is found that the LSTM model heavily exploits the temporal features of the energy consumption data. The forecasting results are also applied in an energy optimization problem to help in identifying the peak demand and smart load scheduling. The findings underscore the fact that ML-based predictive analytics are effective tools that can be used to increase the efficiency of energy, lower the costs of operations, and provide scalable energy optimization in smart buildings. Results indicate that the LSTM-based forecasting model achieved the best performance with a Mean Absolute Error of 8.9 kWh, Root Mean Square Error of 12.3 kWh, and an R2score of 0.93, outperforming linear regression and tree-based ensemble models. Ensemble methods such as Random Forest and Gradient Boosting also demonstrated improved accuracy over baseline regression, confirming the effectiveness of nonlinear learning approaches. These results validate the robustness of deep learning models for capturing complex temporal consumption patterns and their applicability in real-world smart building energy optimization scenarios.